Within referral guidelines for suspected cancer, recommendations for further investigation or referral (henceforth simplified to ‘investigation’) frequently depend on whether symptoms are ‘unexplained’. In its guidance the National Institute for Health and Care Excellence (NICE) defines ‘unexplained’ as:
‘… symptoms or signs that have not led to a diagnosis being made by the healthcare professional in primary care after initial assessment (including history, examination and any primary care investigations).’
1
Clear-cut diagnoses, particularly on initial presentation, are the exception. We cannot dismiss the possibility of cancer just because it is not the most likely diagnosis. Instead, where a cancer explanation exists, we generally accept the much more likely non-cancer explanation only if the likelihood of cancer is very low.2
How to decide whether symptoms are ‘explained’
This article considers the challenge of deciding which symptoms should be considered ‘unexplained’ with respect to a possible cancer diagnosis, as referenced in referral guidelines. Presentations that clinicians determine are not due to a conventional medical diagnosis, which have historically sometimes been termed ‘medically unexplained symptoms’, are not considered.
For several ‘red flag’ cancer symptoms, guidelines recommend referral with little or no qualification. For example, NICE guidelines recommend urgent referral for possible pancreatic cancer in those aged ≥40 years with jaundice. However, guidance for several other symptoms depends on whether the symptoms are ‘unexplained’ (Supplementary Table S1).2 In our experience, various factors may influence clinicians’ preference for labelling symptoms as explained or unexplained. For example, the presence or absence of risk factors may nudge clinicians towards either rejecting or accepting a non-cancer explanation, versus a cancer explanation for particular symptoms. However, factors influencing decision making are not always consciously acknowledged, or explicitly articulated, in a consultation. Table 1 outlines examples of circumstances that might affect decision making. This is intended to facilitate reflection, and does not serve as guidance or judgement as to which presenting circumstances should be designated ‘explained’ or ‘unexplained’. Supplementary Box S12 summarises evidence as to whether comorbidities help to promote or hinder earlier diagnosis.
Table 1. Factors that, in the view of the authors, could affect whether a clinician considers symptoms to be ‘explained’ (that is, cancer not suspected) or ‘unexplained’ (that is, cancer suspected) in situations of uncertaintya
Assessing risk
The positive predictive value for cancer for almost all symptoms highlighted in referral guidelines is less than 5%.1 However, doctors make diagnoses and investigate not only based on the probability, but are also informed by the imperative of identifying serious disease even when other explanations are much more likely (Supplementary Table S1).2 It seems likely that the expectation that cancer and other important diseases are diagnosed promptly has increased and doctors may be mindful of the criticism they may receive in retrospect if diagnoses are perceived to have been ‘missed’ or ‘delayed’. Such pressures may make it harder for doctors to use the traditional ‘test of time’, and they may now be more inclined to arrange investigation or referral for presentations at lower levels of risk.
Such evolving patterns of diagnostic behaviour may be contributing to the changes observed in symptoms leading to cancer diagnosis, with some text-book cancer presentations such as haemoptysis for lung cancer becoming much less common.5 A common symptom that fits a pre-existing diagnosis, such as cough with chronic obstructive pulmonary disease (COPD), does mean that the COPD should be favoured as the likely explanation for the symptom. Even so, the presence of a pre-existing illness does not mean that the individual’s risk of having the cancer is lower. Indeed, individuals with conditions such as COPD have a higher risk of cancer, not only because of the shared risk factor of smoking, but also because of the damage to lung tissue from the COPD itself.6 Patients with chronic disease often recognise their usual symptoms.7 Therefore, it is important for clinicians to ask ‘what has changed and when?’ and to value patients’ own concerns and intuition as a prompt to further action.
Particular caution should be taken in using non-specific tests, such as inflammatory markers ‘to check that nothing is wrong’. The limited sensitivity of such tests means that a negative result is insufficient to rule out cancers. Furthermore, the patients for whom clinicians choose to arrange blood tests have a higher risk of cancer than the general population even after a normal result.8
We offer the following advice in helping to decide whether to consider potential cancer symptoms explained or unexplained.
Share decision making but avoid ‘passing the buck’
Patients’ own preferences and values should inform decision making, but many patients will want their doctor to take the lead on deciding next steps, especially with regards to diagnosis. In our experience openness about excluding cancer as a reason for investigations or referral may not always be welcomed, but we consider that shared decision making relies on patients being fully informed and accepting the responsibilities as well as the rights that stem from a partnership approach.9
Doctors need to strike a balance between sharing decision making, while ensuring that patients who do want clear guidance are not left without recommendations about what action should follow. Where the most appropriate course of action is not clear, clinicians should acknowledge this explicitly. If you are considering specific tests, discuss these with your patient to gauge the acceptability of these along with understanding how concerned your patient is and to what extent they are prepared to tolerate the potential hassle and stress of additional investigations versus the (often low) risk that there is a serious cause for their symptoms.
Recognise biases and diagnostic pitfalls
Symptoms are often already framed within explanations given by patients, colleagues who have documented consultations about the same problem, or our own rapid diagnostic formulations. Anchoring bias is our tendency to maintain such framed explanations, so we may seek evidence to validate such explanations and downplay evidence to the contrary. Mitigating this bias requires doctors to cultivate the discipline of reviewing and challenging diagnoses balancing supporting and conflicting evidence. In some cases it may be appropriate to withhold a diagnostic label altogether to avoid ‘premature closure’.10 Availability bias also matters: we have a tendency for our diagnostic horizons to be limited by the conditions we commonly encounter.
Recognise and mitigate the risk of bias against particular groups
Societal biases adversely affect populations including non-majority ethnic groups. It is important that doctors do not assume that they are not affected by such biases. Strategies to mitigate the risk of misguided reasoning caused by these biases have been described elsewhere.11 These include nurturing genuine curiosity about patients as individuals by seeking to find out about their lives. It may be helpful to take a moment to reflect on whether different diagnoses would be being considered if the individual was a member of a different population group.
Mitigate the hazards of discontinuity …
Discontinuity is increasingly the norm in primary care. It brings hazards, particularly if clinical tasks are taken at face value. For example, a patient may describe their attendance as ‘to get my test results’. The simplest response can be to reassure that the tests were normal. However, this response omits considering why the test was undertaken (indication), the degree of concern (pre-test probability), and the patient’s current symptomatology and concerns. Ideally, clinicians would ‘think ahead’ by documenting the indications for tests along with next steps if these are normal. Such steps may include re-evaluation to consider further investigation if symptoms have persisted or changed (‘safety netting’).
… but embrace the opportunity to bring a fresh perspective
Continuity of care is generally preferred by patients,12 and may particularly benefit those at most risk of cancer, such as older people and those with comorbidities.13,14 However, discontinuity could sometimes promote earlier cancer diagnosis, perhaps by offering a fresh perspective or enabling clinicians to displace anchoring biases by allowing a different clinician to evaluate evidence from scratch.15 Tackling a problem being investigated by another colleague may be uncomfortable for the patient and doctor, but it does offer the opportunity to decide whether explanations attributed to symptoms are satisfactory, or whether further investigation is warranted.
Make use of non-invasive and low-cost cancer tests
Many patients favour testing for cancer in response to symptoms, even where the risk is low.16 However, patients who see their doctors with symptoms frequently do not always receive testing recommended by guidelines.17 At the same time, it is known that overall testing in general practice has increased considerably. A large proportion of non-specific tests that are probably undertaken for vague or misguided indications (for example, ‘to check that nothing is wrong’) are not supported by evidence-based guidelines. In the face of complicated patterns of both under- and over-testing, GPs have a challenging task of testing purposively in response to particular presentations and eschewing the temptation to undertake impulsive and broad-brush testing strategies (‘bloods and review”’) as a mechanism to conclude consultations or defer decision making.
NICE recommends that several tests, such as chest X-ray, cancer antigen 125, and faecal immunochemical test (FIT), are offered for numerous common and often non-specific symptoms.1 These tests are relatively inexpensive, widely accessible, and have well-understood and good performance characteristics. GPs should readily use these tests where recommended by guidelines for particular presentations.
Remember that excluding specific diseases does not exclude other serious illness
There is substantial overlap between several symptoms for different cancers, and between cancers and other serious non-malignant diseases. For example, weight loss is a possible feature of many serious illnesses, including cancer and non-cancer pathologies. Therefore, do not assume that all is well just because tests for specific cancers are negative.
Calibrate your interpretation of test results (post-test probability) with reference to pre-test probability and test performance
Estimation of pre-test probability, namely the probability of a condition existing before we know the results of any diagnostic test, would ideally be informed by known risk of cancer from symptoms and factors such as age, sex, and smoking status. In many circumstances ascertaining pre-test probability from high-quality evidence will not be possible and the clinician needs to make a judgement as to their degree of concern of a serious cause for symptoms in general, or of a specific pathology in particular. The degree of reassurance available from a normal test result should depend upon the pre-test probability, or, in simple terms, how concerned the doctor was when requesting the test coupled with the known performance of the test. For example, it is reasonable to take a normal chest X-ray result as sufficient support for an explanation of viral bronchitis for a cough with no other cancer-related features, whereas a negative FIT, in the context of persistent symptoms plus weight loss, warrants consideration of further testing. Additional resources that may be helpful to further understanding of pre- and post-test probability are listed in Supplementary Box S2.2
If you remain concerned, plan follow-up or use safety netting to empower patients to act
Arrangements for follow-up should be made for patients with ongoing concern, ideally with the same clinician, to decide on any further investigation. This is to reduce the risk of delayed cancer presentations. Safety-netting practices can include informing patients about the expected course of symptoms and can legitimise re-consultation, for example, if symptoms do not resolve as expected.18 However, vague safety-netting advice can be counter-productive or perceived as dismissive.19 Simply informing patients that they will need to obtain a further appointment cannot ignore the difficulties many patients currently have in accessing primary care. Therefore, safety netting should be specific (for example, ‘Book an appointment if it’s not improved in 4 weeks’), and the clinician may have to offer ways to bypass barriers (‘If you can’t get an appointment, then leave a message for me and I’ll call you back’). As well as the practice of individual clinicians, agreed approaches within a clinical team and system-level initiatives are required to maximise safety.20
